Title: Evaluating Automatic Warning Cues for Visual Search in Vascular Images
Author: Boris W. van Schooten, Betsy M.A.G. van Dijk, Anton Nijholt, and Johan H.C. Reiber.
Summary: The authors of this paper focused their study on computer aided visual search. Visual search is a task that traditionally has required humans to implement well, such as monitoring a security camera feed, or analyzing medical scans.
Rather than making a judgement call of the content in a photo or video, the researchers decided to make a system which alerts the human analyzing the picture to possible areas of interest, letting them decide if the area is of concern or not.
The researchers found that users generally prefer paranoid alerts, which display false positives, instead of a conservative system which displays false negatives. This is contrary to earlier reviews, but the performance gains that the researchers discovered are hard to ignore.
Discussion: These IUI papers are much more technical than the ones in the past, but this paper was easy enough to understand. I'm surprised that researchers in the past believed that users would want to see false negatives over false positives. The researcher's findings seem fairly obvious after reading through their paper.
I agree exactly with your sentiments regarding what the researchers believed users would want. Testing for things like cancer, no one wants false negatives.
ReplyDelete